The Research On Feature Points Extraction From Skull Based On Statical Methods

Feature points extraction in three-dimensional model is hot fields in computer vision and pattern recognition. They are widely used in craniofacial restoration, three-dimensional reconstruction, three-dimensional search, three-dimensional registration and other fields. However, the current research focuses on interactive feature point's extraction method which is not conducive to the development of related research. In this paper, we propose a skull feature points extraction algorithm based on statistical method which can be applied to three-dimensional grid of skull model,and then improved the relative angle histogram method and used Bayesian classifier to extract feature points. Design and implement a three-dimensional grid model skull feature point's extraction system.The algorithm can be applied to craniofacial reconstruction based on statistical method. The main contents of this paper are listed as follow:1. The traditional algorithms can not extract the relative angle-context histogram distribution(RACD) feature points of the skull. We improved the RACD in this paper, Experimental result shows that the algorithm can determine the location of the feature points efficiently.2. The points' distribution on skull is uneven, so the number of pre-match collection of points is not easy to determine. We proposed a method based on statistical to estimate the error range, which can dynamically determine the number of pre-match collection on skull model. 3. The local geometry extracting of model. The point's normal is acquired through the model. Based on the skull morphology, we use the normal vector projection in the axis as the properties of data classification.4. Classification based on the principle of none parameter estimation of Bayesian is designed to located the precise feature points. First, the pre-match collection denoising preprocessing, and then located the precise feature points from pre-matches points by classification.The result shows that the accuracy of feature points has been greatly improved.5. Design and implement system of skull feature points extraction. In this paper, a system of three-dimensional grid model of skull feature points extraction is designed and implemented by using matlab toolbox.